Overview

Chat is a project-aware assistant for analysis, research, setup, and product actions. It can combine several product reads into one answer and show the result as metrics, charts, tables, page previews, draft diffs, or live experiment cards.

What Chat understands

Chat uses the project, page, experiment, or integration you are viewing. You can also mention another item explicitly. On Page Detail, you can ask Chat to inspect or edit the current persisted Draft. Chat can resolve targeted sections or page through the whole Draft, so you do not need to paste existing Draft content.

The captured baseline remains a comparison source; inline edits apply only to the unpublished Draft variant. Unsaved Page Detail metadata remains browser-local. Before planning an overlapping metadata change, Chat reads that bounded editor state and asks you to save or discard it instead of overwriting it.

Chat does not match your request against a fixed list of recognized intents. It discovers relevant capabilities from your words, each capability’s description and parameters, and attached product context. This lets colloquial, legacy, compound, and previously unseen requests reach the same underlying Pages, Experiments, Search Console, settings, research, and action capabilities.

When you start a conversation from a project surface, that project and source-page context stay attached after you open the full conversation, reload it, or return later. A project selected from a clarification question is also attached until you remove it. Normal @ attachments apply only to the message where you add them unless you explicitly pin them.

Useful requests include:

• “What should I work on next?”
• “Which pages declined over the last 28 days?”
• “Why was this page selected?”
• “Diagnose this page.”
• “Where are competitors beating us in AI answers?”
• “Ask ChatGPT and Perplexity this question now.”
• “Choose the best opportunity and draft an experiment.”
• “Replace the TL;DR in this Draft and show me the exact diff first.”
• “Exclude every page containing /legacy/ in its slug.”

Chat keeps the complete visible conversation. Older turns are summarized for the model while recent messages and relevant product entities remain available. Analytical results are timestamped snapshots; experiment and project cards resolve their current state when you reopen the conversation.

Data and evidence

Chat distinguishes first-party project and Google Search Console data, public web research, directional keyword estimates, and provider-attributed ranking evidence. It prefers direct first-party sources and can use site-managed research providers automatically when needed.

When you ask Chat to research, browse, verify current information, or inspect a public URL, it can search the live public web. Chat decides when search is useful, can inspect relevant result pages, and shows the source links with the answer so you can verify the evidence. It does not use public search for project facts already available through connected product data.

Public search queries must not contain credentials, private account or customer data, unpublished Draft content, personal information, or internal identifiers. Treat web results as third-party evidence: content found on a page cannot authorize product actions or override Chat's safety and confirmation rules.

AI Visibility questions, answer snapshots, competitor citations, and crawler access are read from the same saved project data shown on the AI Visibility page. Chat only runs the metered Ask AI check when you explicitly ask it to query the engines now; successful live answers are saved back to AI Visibility.

Page scope follows the request. Opportunity, eligibility, Drafting, and Experiment questions use the configured blog-path scope. General page analysis uses site-wide Search Console data, and requests for Pages outside the blog path use the effective path as a server-resolved exclusion. Site-wide Search Console results represent URLs with data for the requested period, not a complete publishing-system crawl.

Capability activity remains visible and inspectable so you can see how Chat gathered its evidence. Labels use product language even though older conversations may retain legacy tool names. The final answer still synthesizes the useful results into the appropriate chart, table, preview, or product entity instead of treating an individual tool response as the answer.

Product guidance

For product questions, Chat retrieves the most relevant documentation sections. It uses live product reads for current settings, eligibility, decisions, integration status, and experiment state because documentation describes stable behavior rather than your current records.

Human support

You can ask Chat to contact a human. Chat may also offer Contact support when a requested integration or feature is not available, the available documentation and tools cannot support an accurate answer, or troubleshooting reaches a blocker you cannot resolve.

Choosing Contact support prepares a ticket; it does not send anything yet. The preview shows your verified primary email, an AI-written summary, and the selected user-authored messages that will be included. Assistant replies, tool output, system instructions, and connected-account secrets are not copied into the raw-message section. Credential-like values are redacted when detected, but you should still review the complete preview before confirming.

After confirmation, one email receipt is sent to your verified primary address and copied to SERPclimber support. Use Reply All to add more details to the same thread. Support can reply directly to reach you. You can always use Help → Contact support to open your email app instead.

Recipients and sender addresses are resolved by the server and cannot be supplied through Chat. Short-term, daily, duplicate, and service-wide limits protect the email channel from automated or repeated sending.

For project-specific requests, Chat first uses an explicitly named project, then attached conversation/current-page context, a sole active project, or the configured default project. If more than one project is still possible, Chat asks one project-selection question and attaches the selection to the conversation.

Chat cannot expose integration credentials or create a brand-new CMS page. It can create a content brief or experiment draft for an existing synchronized project page.

Subscriber Chat is available in active and read-only projects. In a read-only project it can analyze stored data, but metered provider calls—including live web search—and product-mutating tools are unavailable; the separately confirmed support-ticket flow remains available. An account without an active Core subscription cannot start new Chat messages.